AIM: To describe the problems in the interpretation of Croatian mortality data and explore possible reasons for the recorded increase in mortality in the 1990-1999 period, particularly related to different methods of collection and estimation of data on deaths and population. METHODS: Numbers of recorded deaths and population estimates were first obtained from the Croatian Institute for Public Health and examined in detail. The Institute used population estimates supplied by the Croatian Statistics Bureau, which included de jure population data (including all Croatian citizens wherever they live) until 1996 and de facto population data (including only population living in Croatia at least for a year, irrespective of citizenship) since 1996. A different set of population estimates based on de facto estimates since 1992 was obtained from the Croatian Bureau of Statistics. We examined trends in age- and sex-specific death rates from major causes in 1990-1999 period, using the mortality data from the Croatian Institute for Public Health and both sets of population estimates. Lung cancer as a cause of death was examined in more detail, since it is relatively stable over short periods of time. Interviews were undertaken with key informants to identify the reasons for any discrepancies. RESULTS: In Croatia, relatively stable death rates from lung cancer in men ranged from 84/100,000 in 1990 to 79/ 100,000 in 1995. In 1996, a marked discontinuity appeared in the Croatian data, with a 14% increase compared to 1995 (from 79/100,000 to 91/100,000) and a further increase in 1999 (94/100,000), which is not credible on the basis of the natural history of lung cancer. Analysis of mortality rates with de facto population estimates showed more gradual increase from 1992-1996. Methods used to estimate population and mortality during the 1990s were inconsistent and misleading. At present, it is impossible to be certain about the true level of mortality in Croatia during 1990s, as the numerator (deaths) and denominator (population) were incompatible until 1998. CONCLUSION: Major problems in data collection would have been identified if the investigation of unexpected mortality trends in Croatia in the 1990s had been done. Systematic analysis of health patterns should be done as soon as data from the 2001 census become available. Capacities in public health should be strengthened to make this possible. This issue has received little recognition from the international donor organizations, particularly those that use health data.